[Can data from quality assurance programs such as peri-/neonatal compilations be used for secondary epidemiological...

The aim of this study was to explore whether population-based data from a regional quality control program can be utilized to compare the neonatal outcome of small for gestational age (SGA) and appropriate size for gestational age (AGA) new-borns.

The Center for Quality Management in Health Care maintains perinatal data for almost all births in Lower Saxony (Germany). Neonatal data are collected for all infants admitted to hospital within 10 postnatal days. We evaluated linked perinatal and neonatal datasets of 4126 very low birthweight infants (VLBW; < 1500 g), born in 1991 - 1996. After checking for completeness, representativeness, and validity, exclusion criteria were defined to minimize bias and to yield similar proportions of SGA- and AGA-neonates. Since inclusion of all multiple births would lead to an overestimation of maternal risk factors, one sibling was randomly selected from each set of multiples. Bias arising from not well defined study populations should be shown based on univariable mortality analyses (Kaplan-Meier survival curves).

Application of exclusion criteria resulted in a final study population of 1623 independent (disjunctive) new-borns from 25 - 29 weeks gestation, 173 of whom were SGA, 1450 AGA. Kaplan-Meier curves from the initial study population and the well defined study base differed significantly. Trend analysis revealed a significant (p < 0.05) increase in proportions of both VLBW (from 0.95 % in 1991 to 1.11 % in 1996; + 17 %) and SGA infants (from 22.7 % to 27.4 %; + 21 %) within the observational period. A well defined data selection process is necessary if data collected for other purposes are to be used for epidemiological studies. Neglecting this labour-intensive work may be one reason for the varying results on the outcome of SGA and AGA infants.